Factor and figure management; Repo hygiene

Yidie Feng
10/12/2017

library(gapminder)
library(dplyr)
library(ggplot2)
library(tidyverse)
library(forcats)
library(RColorBrewer)

Factor Management (Gapminder Version)

• Remove observations associated with Oceania and drop the unused level.

gapminder2 <- gapminder %>%
  filter(continent != "Oceania") %>%
  droplevels()

• Now we investigate the data before and after removing the rows associated with Oceania.

Before, there are five levels of factor continent: "Africa", "Americas", "Asia", "Europe" and "Oceania". After, there are four levels of factor continent: "Africa", "Americas", "Asia" and "Europe".

levels(gapminder$continent)

## [1] "Africa"   "Americas" "Asia"     "Europe"   "Oceania"

nlevels(gapminder$continent)

## [1] 5

levels(gapminder2$continent)

## [1] "Africa"   "Americas" "Asia"     "Europe"

nlevels(gapminder2$continent)

## [1] 4

Before, there are 1704 rows. After, there are 1680 rows.

nrow(gapminder)

## [1] 1704

nrow(gapminder2)

## [1] 1680

• Reorder the levels of country or continent. I will reorder the levels of continent according to mean of lifeExp in descending order.

fct_reorder(gapminder$continent, gapminder$lifeExp, mean, .desc = TRUE) %>% 
  levels() %>% head()

## [1] "Oceania"  "Europe"   "Americas" "Asia"     "Africa"

• Explore the effects of arrange(). Does merely arranging the data have any effect on, say, a figure? My answer is no by looking at the following two plots. After I arrange() gapminder by lifeExp and then plot, it is the same as the plot without arranging. So merely arranging the data does not affect a figure. By looking at the output of levels() function, we can see that the levels of country stay the same after arranging.

This is a subset of gapminder which has the data for Americas in 2002.

gap_america_2002 <- gapminder %>% filter(year == 2002, continent == "Americas")

This is a plot of lifeExp for all the countries in gap_america_2002.

gap_america_2002 %>%
  ggplot(aes(x = lifeExp, y = country)) + 
  geom_point() +
  labs(title="Life Expectancy of Countries in Americas in 2002")

head(gap_america_2002)

## # A tibble: 6 × 6
##     country continent  year lifeExp       pop gdpPercap
##      <fctr>    <fctr> <int>   <dbl>     <int>     <dbl>
## 1 Argentina  Americas  2002  74.340  38331121  8797.641
## 2   Bolivia  Americas  2002  63.883   8445134  3413.263
## 3    Brazil  Americas  2002  71.006 179914212  8131.213
## 4    Canada  Americas  2002  79.770  31902268 33328.965
## 5     Chile  Americas  2002  77.860  15497046 10778.784
## 6  Colombia  Americas  2002  71.682  41008227  5755.260

head(levels(gap_america_2002$country))

## [1] "Afghanistan" "Albania"     "Algeria"     "Angola"      "Argentina"  
## [6] "Australia"

We arrange the data by lifeExp and then plot.

a<- gap_america_2002 %>%
  arrange(lifeExp)
  
ggplot(a, aes(x = lifeExp, y = country)) +
  geom_point() +
  labs(title="Life Expectancy of Countries in Americas in 2002") 

head(levels(a$country))

## [1] "Afghanistan" "Albania"     "Algeria"     "Angola"      "Argentina"  
## [6] "Australia"

• Explore the effects of reordering a factor and factor reordering coupled with arrange(). Especially, what effect does this have on a figure? From the following plots, we can see that reorder() does have an effect on how the figure looks. After I reorder the levels of country by lifeExp and then plot, the countries on the y-axis appear to have order from lowest lifeExp to highest lifeExp when viewed from bottom to top. And reorder() coupled with arrange() gives us the same plot as only using reorder(), since arrange() does not really affect the figure, as explored in the previous question.

We reorder the levels of country by lifeExp and then plot.

gap_america_2002 %>%
  ggplot(aes(x = lifeExp, y = fct_reorder(country, lifeExp))) +
  geom_point() +
  labs(title="Life Expectancy of Countries in Americas in 2002")

We reorder coupled with arrange the levels of country by lifeExp and then plot.

gap_america_2002 %>%
  arrange(lifeExp) %>%
  ggplot(aes(x = lifeExp, y = fct_reorder(country, lifeExp))) +
  geom_point() +
  labs(title="Life Expectancy of Countries in Americas in 2002")

File I/O

Bring rectangular data in using read_tsv()

gap_tsv <- system.file("gapminder.tsv", package = "gapminder")
gapminder <- read_tsv(gap_tsv)

## Parsed with column specification:
## cols(
##   country = col_character(),
##   continent = col_character(),
##   year = col_integer(),
##   lifeExp = col_double(),
##   pop = col_integer(),
##   gdpPercap = col_double()
## )

# change continent and country to factors
gapminder <- gapminder %>%
  mutate(country = factor(country),
         continent = factor(continent))

Compute something worthy of export. Here I will compute the mean of gdpPercap for each country.

gap_mean_gdp <- gapminder %>%
  group_by(country) %>%
  summarise(mean_gdp = mean(gdpPercap))

Then I will write gap_mean_gdp out to get a comma-delimited file. Now we have a file called "gap_mean_gdp.csv" saved in the current working directory.

write_csv(gap_mean_gdp, "gap_mean_gdp.csv")
gap_via_csv <- read_csv("gap_mean_gdp.csv")

## Parsed with column specification:
## cols(
##   country = col_character(),
##   mean_gdp = col_double()
## )

gap_via_csv$country <- as_factor(gap_via_csv$country)

Reorder the levels of country in increasing order of mean gdp. We can see that the levels of country are changed (starting from "Myanmar").

head(levels(gap_mean_gdp$country))

## [1] "Afghanistan" "Albania"     "Algeria"     "Angola"      "Argentina"  
## [6] "Australia"

gap_mean_gdp <- gap_mean_gdp %>% 
  mutate(country = fct_reorder(country, mean_gdp))

head(levels(gap_mean_gdp$country))

## [1] "Myanmar"    "Burundi"    "Ethiopia"   "Eritrea"    "Mozambique"
## [6] "Malawi"

Note that the ordered country levels are maintained withsaveRDS()/readRDS()

saveRDS(gap_mean_gdp, "gap_mean_gdp.rds") 
gap_via_rds<-readRDS("gap_mean_gdp.rds")
head(levels(gap_via_rds$country))  

## [1] "Myanmar"    "Burundi"    "Ethiopia"   "Eritrea"    "Mozambique"
## [6] "Malawi"

dput() creates an R-specific-but-not-binary representation.

dput(gap_via_rds, "gap_mean_gdp_dput.txt")

We can look at "gap_mean_gdp_dput.txt" to get a sense of what it looks like. The function dget() reads the representation back in to our usual data form.

gap_via_dget<-dget( "gap_mean_gdp_dput.txt")

I will summarize in a table whether the country levels survive through these different write/read functions.

This is the country levels stored in gap_mean_gdp which has been reordered.

country_levels <- tibble(original = head(levels(gap_mean_gdp$country)))
country_levels

## # A tibble: 6 × 1
##     original
##        <chr>
## 1    Myanmar
## 2    Burundi
## 3   Ethiopia
## 4    Eritrea
## 5 Mozambique
## 6     Malawi

Add columns for country levels stored in each of .csv/.rds/dput.txt files. We can see that csv gives us alphabetic order. While rds and dput can maintain the ordered levels.

country_levels<-country_levels %>%
  mutate(levels_via_csv = head(levels(gap_via_csv$country)),
         levels_via_rds = head(levels(gap_via_rds$country)),
         levels_via_dput = head(levels(gap_via_dget$country)))

knitr::kable(country_levels)

Visualization design

I will make two figure using what we learned in recent classes.

p1<- gapminder %>%
  filter(year==2007, pop>80000)%>%
  ggplot(aes(gdpPercap, lifeExp)) +
  geom_point(aes(size=pop, fill=country), shape=24) +
  theme_bw() +
  facet_wrap(~continent) +
  scale_color_manual() +
  guides(fill=F)

p1

Writing figures to file

Show a situation where argument p matters. Create another figure

p2<-gapminder %>%
  filter(country == "China"|country == "Canada") %>%
  ggplot(aes(x=lifeExp)) + 
  geom_density(aes(group=country, colour=country, fill=country), alpha=0.3) 

If I want to save p1, I will need to specify "plot=" argument in ggsave() because by default it will save the last plot display which is p2. There are also arguments like width/height/dpi that we can specify.

ggsave("p2.png", width=40, height=40, dpi=400, )
ggsave("p1.png", plot=p1)

## Saving 7 x 5 in image

Embed the two figures:

NOTE: Process report is here.